stat¶
Parameters¶
Parameter label |
I/O type |
Data type |
Mandatory parameter? |
Default value |
---|---|---|---|---|
|
input |
vector |
yes |
N/A |
|
input |
string |
no |
|
|
input |
bool |
no |
|
|
input |
bool |
no |
|
|
input |
integer |
no |
|
|
input |
integer |
no |
|
|
output |
vector |
N/A |
N/A |
Functionality¶
Module can compute the following statistical descriptors from inputVec
based on the value of measure
:
Minimum (
"min"
)Maximum (
"max"
)Range (
"range"
) : max - minMean (
"mean"
)Variance (
"var"
)Standard deviation (
"std"
)Median (
"median"
)Entropy (
"entropy"
): Entropy of raw data vectorEntropy (
"entropy_hist"
): Entropy of a histogramZipf coefficient (
"zipf_coefficient"
)Circular mean (angle) (
"circ_mean_angle"
)Circular mean (length) (
circ_mean_length"
)Circular variance (
circ_var"
)Circular standard deviation (
circ_std"
)Circular dispersion (
circ_disp"
)
If normalizeEntropy=True
, the entropy is normalized to yield values between 0
and 1
, by dividing by log N
with is the maximum entropy for N
classes. N
can be
given either explicitly by the parameter numberClasses
or will be implicitly set as the observed number of different classes. The parameter normalizeToDensity
is useful only for
the entropy_hist
-measure, which normalises raw bin counts to densities. Circular (directional) statististics are need for transformations such as pitch classes (i.e., derived from some
modulo operation), see https://en.wikipedia.org/wiki/Directional_statistics for more details. For circular statistics, the circ_max
parameter is used to map raw values onto the
interval [0, ]. The Zipf coefficient is the (negative) slope of a linear regression on the log-log-plot of rank-ordered frequencies. It is the exponent of a fitted power (Zipf) law
(see https://en.wikipedia.org/wiki/Zipf’s_law).